Back to skill

Security audit

Knowledge Graph Buil

Security checks for vulnerabilities and agentic risk

Overview

This skill is a broad knowledge-graph helper with disclosed local storage, query, visualization, migration, and integration features, but users should review its sync and callback settings before use.

Before installing, confirm you want a skill that can write persistent knowledge-graph files and run kg/Bash examples. Back up existing memory/knowledge-graph data before migration or restore commands, review config files before overwriting them, and enable callback/event/sync integrations only for trusted endpoints and subscribers.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The primary description is written predominantly in Chinese with only partial English fragments, and the document does not state that other languages are supported or let the user opt into a locale. This can violate language/locale policy where skills should not force a specific language without user choice.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

Cross-skill communication, callbacks, and event bus features are described without clear privacy boundaries or warnings that graph/task data may be transmitted to other skills or remote endpoints. Because this skill handles structured knowledge and project/task metadata, implicit sharing can expose sensitive internal relationships, plans, or identifiers beyond the original trust scope.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The setup examples create directories and overwrite configuration files in persistent storage using redirection without warning the user that local data and configs will be modified. In an agent environment with Write/Edit/Bash permissions, this can silently alter state, break existing setups, or replace trusted configuration with attacker-influenced content.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The migration command changes the storage backend and persistent data layout but is presented as a routine operation without warning about stateful side effects, rollback requirements, or integrity checks. An agent following this documentation may irreversibly alter storage, causing data corruption, downtime, or compatibility issues if run on the wrong dataset.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill claims commands do not concatenate user input, yet examples such as deploy labels and JSON props interpolate shell variables directly. In an agent setting, operators may copy these patterns into Bash execution paths, enabling command injection or unsafe argument expansion if variables derive from untrusted workflow data.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The metadata frames the skill as a local knowledge-graph/database tool, but the body also advertises external API integration, HTTPS callbacks, CI/CD sync, and cross-platform communication. This scope expansion can cause an orchestrating agent or user to grant network-capable use in contexts that expected only local storage operations, increasing exfiltration and unintended outbound access risk.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The documented callback_url and event/cross-skill features introduce data transmission outside the local graph workflow without a clear necessity statement or restrictions on what may be sent. In agent systems, ambiguous callback support can be abused to route graph contents, task metadata, or status events to attacker-controlled endpoints.

Content

No source excerpt is available for this finding.